Practical schemes to utilize multiple antennas in wireless communications were initially directed towards achieving diversity gain (see Naguib, Tarokh, Seshadri and Calderbank, A space-time coding modem for high-data-rate wireless communications, IEEE JSAC, October 1998, and Alamouti, A simple transmitter diversity technique for wireless communications, IEEE JSAC 1998). Space-time coding was concentrated mainly on achieving the best possible performance using only the channel distribution known at the transmitter 12. In a parallel manner, the assumption that the instantaneous channel information is available at the transmitter resulted in work in transmission beamforming and in antenna selection (see: Farrokhi, Liu and Tassiulas, Transmit beamforming and power control for cellular wireless systems, IEEE JSAC, October 1998, Gelrach and Paulraj, Adaptive transmitting antenna methods for multipath environments, Globecom 1994, and Winters, Switched diversity with feedback for DPSK mobile radio systems, IEEE Tran. Veh. Tech., February 1983). These methods dealt with transmissions that have a channel rate of one symbol per channel or lower, and all of these diversity methods can be used even when there is only one receiver antenna. When multiple receiver antennas are available, these antennas were simply used to add receiver diversity gain.
The achievable capacity of MIMO wireless communications in the presence of multiple transmitter and receiver antennas was perceived to be much larger, as described by Telatar (Capacity of multi-antenna gaussian channels, Bell Systems Technical Journal, 1995) and by Fochsini and Gans (On the limits of wireless communications in a fading environment when using multiple antennas, Wireless Personal Communications, 1998). A step towards achieving this capacity was taken by the introduction of the layered space-time architecture, also known as diagonal BLAST, which explains how the multi-dimensional channel can be used to deliver several one-dimensional streams of data, in an environment where the channel state information (instantaneous Rayleigh fading channel value) is known at the receiver, but not at the transmitter (Varanasi and Guess, Optimum decision feedback multiuser equalization with successive decoding achieves the total capacity of the Gaussian multiple-access channel, Asilomar 1998). Vertical Bell Labs Layered Space-Time Code (V-BLAST), which is a simpler implementation, advocates a simple demultiplexing of the data stream instead of some specific encoding in space-time (see Wolniansky, Fochsini, Golden and Valenzuela, V-BLAST: An architecture for realizing very high data rates over the rich-scattering wireless channels, Signals, Systems, and Electronics, 1998. ISSSE 98. 1998 URSI International Symposium, September 1998). The corresponding receiver architecture for V-BLAST is also simpler (see Fochsini, Reinaldo, Valenzuela and Wolniansky, Simplified processing for high spectral efficiency wireless communications employing multi-element arrays, IEEE JSAC, November 1999). A step closer towards achieving capacity is taken by assuming the availability of some channel information at the transmitter. The PARC method (see Chung, Lozano and Huang, Approaching eigenmode BLAST channel capacity using V_BLAST with rate and power feedback, VTC Fall 2001) is an example of such a technique. Here, two antennas are allotted variable rates and powers, according to their respective channel conditions. The encoding is done separately on these two streams. In such a situation, the optimal receiver (in a capacity-achieving sense) was discussed by Varanasi and Guess (see Varanasi and Guess, Bandwidth-efficient multiple-access via signal design for decision feedback receivers: Towards an optimal spreading-code trade-off, Globecom 1997, and Varanasi and Guess, Optimum decision feedback multiuser equalization with successive decoding achieves the total capacity of the Gaussian multiple-access channel, Asilomar 1998).
It is pointed out that a further aspect of the foregoing, used in conjunction with rate control of the different streams, is the transmission of streams over eigen-beams rather than over separate antennas.
It should be noted, however, that the availability of channel information cannot be taken for granted, since it involves some reciprocity assumptions in the channel, or some feedback from the receiver. When partial channel knowledge is available, a criterion for switching between diversity and multiplexing was discussed in Heath and Paulraj, Switching between multiplexing and diversity based on constellation distance, Allerton 2000. More recently, multiplexing mechanisms specific to CDMA, which combine code multiplexing with space-time multiplexing, were presented by Huang, Viswanathan and Fochsini, Multiple antennas in cellular CDMA systems: Transmission, detection and spectral efficiency, IEEE T-Wireless, July 2002.
It has been shown in theory that the optimal approach for MIMO transmission, called the eigenmode or water-filling MIMO in the literature, is to transmit multiple streams of data, where the encoding rate and power allocation of each stream is tailored to the channel quality that is seen by each stream. Also, MIMO theory has shown that the best performance is obtained when a data packet is jointly encoded and interleaved across eigen-beams. One possible transmission technique for jointly encoded packets with a systematic code was referred to as Flexible Rate Split (FRS), where the number of systematic bits in each of the eigen-beams is controlled. In a simple yet practical implementation of the FRS algorithm for 2-beam transmission, as many systematic bits as possible are transmitted over the first stream, and the remaining systematic bits, as well as the parity bits, are transmitted over the second stream. It has been found that, for higher, fixed coding rates (>0.5 for 2-beam transmission), this technique produced poor frame error rates, even though the bit error rate performance was very good. The reason for this dichotomy is that, when the strength of the second eigen-beam is very poor (i.e., the channel is almost rank-1), some systematic bits are lost in the second beam, hence giving rise to poor decoding performance.
Related to the foregoing, known types of algorithms that attempt to implement maximum information rate designs typically employ water-filling or water-pouring optimization at the transmitter (part of encoding and modulation), when the channel state information is available to the transmitter. In essence, the water filling approach identifies and accesses sub-channels defined by eigenmodes of a MIMO channel matrix, and then allocates optimal fractions of the total energy available at the transmitter to the available sub-channels. The two main assumptions underlying this approach are that: (1) the encoding and modulation scheme is capable of operating at, or near, information capacity limits, and (2) ergodicity holds, i.e. theoretical ensemble averages are achievable via time averages. The ergodicity insures that the resulting information rate can be attained with arbitrarily low error probability by the coding scheme, in the channel under consideration, via an adequate number of exposures of the coded symbols to sufficient channel realizations. When the first assumption is true, the inherent discarding of one or more sub-channels during water-pouring, as is recognizable by the presence of the (·)+ operator, which replaces a negative argument by 0 in the solution to the water-filling problem, has no sensible impact on performance as long as information capacity is well-defined, and the system remains in the proximity of capacity limits. Reference with regard to the presence of the (·)+ operator can be made to: M. Medard, “The effect upon channel capacity in wireless communications of perfect and imperfect knowledge of the channel,” IEEE Trans. Inform. Theory, vol. IT-46, pp. 933-946, May 2000; R. G. Gallager, Information Theory and Reliable Communication, New York: Wiley, 1968; T. M. Cover, and J. A. Thomas, Elements of Information Theory, New York: Wiley, 1991; and H. Sampath, P. Stoica, and A. Paulraj, “Generalized Linear Precoder and Decoder Design for MIMO Channels Using the Weighted MMSE Criterion,” IEEE Trans. Commun., vol. 49, no. 12, pp. 2198-2206, December 2001.
However, the foregoing assumption is often not true, as the coding scheme can typically operate away from information capacity limits. Regarding the second assumption, it is itself often precluded by the fact that the channel is quasi-static, or encoding is performed without sufficient exposures of the coded symbols to enough channel realizations. Furthermore, in quasi-static scenarios the meaningful information theoretic limit is given by outage capacity, rather than ergodic (Shannon) capacity.
When either of the above assumptions is not met, discarding one or more eigen sub-channels (as a result of implementing some water-filling algorithm) impacts the performance corresponding to that particular realization of the MIMO channel matrix.
The situation becomes more problematic in any MIMO systems (e.g., in a 1XEvDV MIMO system) that use a quasi-complementary approach to (possibly capacity achieving) adaptive coding schemes. One example is a turbo scheme, whereby systematic information must be transmitted (1) in its entirety (i.e., without puncturing), (2) separately from parity check information, and (3) while adaptively controlling the overall code rate (spectral efficiency) by sending all systematic symbols, and only a part of the parity check symbols. In the above scenario, MIMO channels arise from the use of multiple transmit antennas in order to either: provide a spatial dimension to the encoder (e.g., space-time codes); or, to add further spatial redundancy whenever possible, via eigen beamforming or (MIMO) channel precoding.
A goal in 3.5 G and 4 G (respectively, generation three-and-a-half and four) systems is to achieve high data rate at relatively low cost. Throughputs of 1 Gbps (local area) or 10 Mbps (wide area) result in high spectral efficiencies. Physical limitations due to higher propagation losses incurred at the higher carrier frequencies to be used in 4 G systems result in smaller cell areas. Especially in 3.5 G systems, capacity and throughput are at a premium.
As a result, it can be appreciated that efficient resource allocation is crucial in achieving the targeted throughput, while still controlling cost. Bit loading and, in general, controlling the relevant transmission parameters are important elements when attempting to approach the capacity limits, and to thereby use the spectral resources efficiently. All schemes that are known to the inventors for attempting to achieve optimum resource allocation at the transmitter (which presumably has channel state information) require precoding, with the goal of accessing the eigen sub-channels in an optimal manner.
Based on the foregoing, it should be appreciated that a need exists to provide a solution to those operational scenarios that recognize a need to treat some of the coded symbols preferentially, by a QoS guarantee, where an example encompasses the systematic symbols in a turbo coded frame. In this case, the systematic symbols are key to the successful decoding of a particular frame, and are important when it is desired to reduce the number of frame retransmissions.
In MIMO channels that admit eigenmodes, one technique to insure preferential treatment for some of the coded symbols is to mount the critical coded symbols, e.g., the systematic symbols in a turbo code, on the stronger eigen sub-channel(s). However, a problem can arise if the available sub-channels cannot accommodate all of the critical symbols in a frame, and may result in having to de facto drop some of the critical symbols.
It quickly becomes apparent that a pure eigen beamforming approach creates the risk of being unable to transmit all of the critical (e.g., systematic) symbols, should the water-filling approach result in ‘clipping’ the weaker eigen sub-channel, unless, of course, the base station is willing to reduce the throughput for that particular frame in order to allow a resultantly smaller number of systematic symbols per frame to fit entirely on the stronger eigen sub-channel(s).
Generally, conventional approaches to the foregoing problems focus on pure water-filling designs, which attempt to maximize the information rate. However, many current systems do not accommodate operation near the information capacity limits. This can be due to the use of legacy designs, or to short duration frame and slot structures that preclude sufficient channel variation at reasonable rates of change (with respect to channel estimation).
It is noted that some MIMO transmission methods exist for splitting an encoded data packet into multiple streams without differentiating between the streams (Double Space-Time Transmit Diversity (DSTTD), DABBA and Vertical Bell Labs Layered Space-Time Code (V-BLAST). Techniques also exist for splitting the data into multiple packets of different sizes and encoding them separately (Per Antenna Rate Control (PARC)).
However, no adequate solutions are known by the inventors for coping with the preferential treatment of a subset of coded symbols, such as the systematic symbols in a turbo coded frame, during transmission via eigen sub-channels of a MIMO channel.